AI image fraud to cost $40 billion by 2025 - can new standards help?
International efforts aim to authenticate digital imagery amid rising deepfakes
1The credibility crisis in imagery
AI-generated fake images and deepfakes have become so prevalent that distinguishing real photos from synthetic ones is nearly impossible. This creates serious trust issues across various sectors from news media to legal evidence. Deloitte estimates AI-enabled fraud losses could reach $40 billion in the US alone by 2027.
2Role of standards
International standards bodies ISO, IEC and ITU are developing JPEG Trust, a new set of standards that embeds metadata into JPEG files to verify image origins. However, experts warn these standards alone won't stop bad actors, as malicious users won't voluntarily label their AI-generated content.
3JPEG Trust specifications
The new standard consists of two main parts: Part 2 provides trust profile templates for different use cases, while Part 3 introduces watermarking systems for media assets. These systems are designed for applications in broadcasting, digital cameras, and AI content generation.
4Limitations of standards
Project lead Prof. Ebrahimi emphasizes JPEG Trust doesn't label images as 'trusted' or 'untrusted'. Instead, it provides tools for users to make contextual decisions. 'Someone breaking laws won't label their content as AI-generated,' he explains, highlighting the system's inherent limitations against determined fraudsters.
5Fragmented industry efforts
Until now, combatting deepfakes has involved fragmented approaches, with companies developing proprietary solutions while some collaborate on de facto standards. IEC and ISO recognize the urgent need for universally accepted industry standards to address this growing challenge.
6Future outlook
As AI-generated content proliferates, image authentication standards will become increasingly crucial. While initiatives like JPEG Trust represent important steps toward rebuilding digital trust, they likely won't completely prevent misuse. Enhancing users' digital literacy remains equally important in this evolving landscape.
Original article: Latest news — Yapay zeka sahtekarlığı 2025'te 40 milyar dolara mal olacak